bauhaus-visual

Bauhaus Visual

Compile a title, theme, purpose, or interior reference image intooneBauhaus visual prompt that can be handed directly to an image model.

Core principles:First determine the medium, then apply the rules. Bauhaus is not red-yellow-blue geometric decoration; different media must be broken down into different execution priorities: covers focus on layout, illustrations focus on conceptual structure, posters focus on text-image impact, and interior redraws focus on space, furniture, materials, and circulation.

Read

Read in the following order, then begin compiling:

• references/compile-contract.md: compilation contract, field order, output quality threshold. Must be satisfied before every output.
• references/module-library.md: module library for purpose, aspect ratio, key visual, color, text mode, and interior redraw rules.
• references/presets.md: 4 compiled presets. When the user just wants something ready to use or provides very little input, directly choose the closest one and adapt it.

Input

Prioritize inferring from the user's words; do not require form filling:

1. Purpose: article cover, body illustration, column/event poster, interior photo redraw. Infer from context when missing; if it cannot be inferred, ask only this one question.
2. Title / Theme: usually needed for covers and posters; body illustrations can use concepts; interior redraws may not need a title.
3. Aspect ratio: when unspecified, covers default to 4:5, body illustrations default to 16:9, posters default to 2:3, and interior redraws follow the reference image or 16:9.
4. Reference image: if the user uploads an interior photo, prioritize interior redraw and preserve the original image's spatial structure.

If the user actively specifies aspect ratio, main title, color scheme, whether to preserve text, etc., the user's choice prevails; but it must not break style purity or medium boundaries.

Workflow

1. Parse intent: determine the use case:article-cover、inline-illustration、poster、interior-redesign choose one of four.
2. Select modules: according to module-library.md assign one value to each slot: aspect ratio, key visual, color, text mode, material/space rules, prohibitions.
3. Compile prompt: output one English art direction according to the field order in compile-contract.md . Parse out all {}、/ options and Chinese annotations; title text is the exception.
4. Output: one finished prompt + one line of rationale for the selection + one line for the closest preset. Do not output the module menu, reasoning process, self-check list, or multiple version options.

Routing Rules

• Article cover: title area, geometric layout, visual focal point, white space, short text control.
• Body illustration: conceptual explanation, structural diagram, geometric relationships, low noise, little or no text.
• Column/event poster: exhibition poster feel, strong title, integrated text-image, stronger compositional tension.
• Interior photo redraw: Preserve the original image's perspective, proportions, doors and windows, walls, and the positions of major furniture; change only the materials, furniture language, lighting fixtures, colors, and spatial atmosphere.

Output Format

Unless necessary input is missing and a follow-up question is required, output only:

1. The complete finished prompt.
2. 选型理由：...
3. 最接近预设：...

Usage Examples

These requests should directly trigger this skill:

生成一张 4:5 包豪斯风格文章封面，标题是：结构比装饰更重要。

生成一张 16:9 包豪斯风格正文配图，主题是：提示词不是描述风格，而是组织结构。

生成一张 2:3 包豪斯风格活动海报，标题是：FORM FOLLOWS FUNCTION。

参考我上传的室内照片，把这个房间重绘成包豪斯风格工作室。

Quality Gates

• The output must not retain {变量}, slash menus, Chinese comments, or multiple sets of options.
• Choose only one main visual each time; do not list multiple engines side by side.
• Put negative items together in AVOID。
• Do not stuff poster layout rules into interior redrawing, and do not stuff furniture or glass curtain walls into article covers.
• Interior redrawing must clearly preserve the original spatial structure; the reference photo must not be completely redrawn into an unrelated model room.
• Do not fabricate image generation results; this skill only produces prompts.